WEBVTT - Stability AI CEO on ChatGPT Rival 0:00:01.560 --> 0:00:06.840 From Marhart where Innovation, Money and Power Collie in Silicon Valley, NBN. 0:00:07.200 --> 0:00:10.720 This is Bloomberg Technology with Caroline Hyde and Ed. 0:00:10.760 --> 0:00:11.440 Love Love. 0:00:25.200 --> 0:00:27.720 And Caroline Hyde a Bloomberg's Weld headquarters in New York. 0:00:28.120 --> 0:00:31.840 Annahmed Lovelow in San Francisco. This is Bloomberg Technology. 0:00:31.360 --> 0:00:34.760 Coming up a conversation about the future of technology with 0:00:34.800 --> 0:00:37.640 the CEO of Stability AI, the houp startup that has 0:00:37.720 --> 0:00:39.960 just launched a chatbot to rival chat GBT. 0:00:41.080 --> 0:00:44.200 Then Tesla raises prices of some models in the US 0:00:44.240 --> 0:00:47.080 one day after CEO e Long Musk hinted at further 0:00:47.159 --> 0:00:50.800 price cuts. This is more shareholders flag concerns that Musk 0:00:51.280 --> 0:00:52.160 is stretched too thin. 0:00:52.800 --> 0:00:55.480 An Earth Day is coming up and we'll break down 0:00:55.520 --> 0:00:59.280 the pros the cons of carbon hatchare technology as more 0:00:59.320 --> 0:01:03.560 companies to offset emissions. First says check in what's happening 0:01:03.600 --> 0:01:05.800 in terms of public markets. Let's move it on, because 0:01:05.800 --> 0:01:07.600 I wanted to show what was happening in terms of 0:01:07.760 --> 0:01:10.040 some of the benchmarks. I don't want to steal Ed's thunder, 0:01:10.080 --> 0:01:11.839 as I think this is one of his individual stocks 0:01:11.840 --> 0:01:13.880 to watch, but we are seeing at the moment maybe 0:01:13.959 --> 0:01:16.840 completely flat on the Nasdaq. Overall, we're still trying to 0:01:17.080 --> 0:01:21.280 pass some of the overall view on the macro perspective here. 0:01:21.280 --> 0:01:24.640 We actually had the PMI data, so business activity showing strength. 0:01:24.880 --> 0:01:26.840 So does that mean the Federal zerv does have to hike? 0:01:26.880 --> 0:01:28.640 We are flat on the NASDAK. The two year yield 0:01:28.680 --> 0:01:30.800 just interest up four basis points. Has been far more 0:01:30.840 --> 0:01:33.240 volatile than some of the benchmarks in terms of stocks recently, 0:01:33.360 --> 0:01:36.759 I shine a light and crypto actually having its downward week. 0:01:36.800 --> 0:01:38.600 We're off by seven percent on the OG in the 0:01:38.640 --> 0:01:40.959 crypto space, just off by six tents of percent on 0:01:41.000 --> 0:01:43.959 the day. Relatively muted volatility there. Let's moveing on and 0:01:43.959 --> 0:01:47.000 have a little look on what's happening therefore in really 0:01:47.080 --> 0:01:49.200 where we look towards earning z because I think this 0:01:49.240 --> 0:01:51.840 is what is capturing traders' attention and why we're seeing 0:01:51.880 --> 0:01:54.960 some caution in the market. We have overall seen well 0:01:55.000 --> 0:01:57.400 perhaps well we'll move over into the micro. I think 0:01:57.400 --> 0:01:59.000 we can't get that particular chart right now, but I 0:01:59.040 --> 0:02:01.520 wanted to show just how much of the heavy lifting 0:02:01.560 --> 0:02:03.560 technology is done, particularly in the S and P in 0:02:03.560 --> 0:02:04.440 the first quarter ED. 0:02:05.120 --> 0:02:07.280 It's why we continue to track the news cycle because 0:02:07.280 --> 0:02:09.359 you look at points movers on the Nasdaq one hundred, 0:02:09.400 --> 0:02:12.840 that is what is driving indexes in either direction right now. Amazon, 0:02:12.880 --> 0:02:15.120 as you said, I was quite excited about this because 0:02:15.160 --> 0:02:15.880 it's moving through. 0:02:15.760 --> 0:02:17.360 The upside, up more than three percent. 0:02:17.680 --> 0:02:19.799 The headline of the last twenty four hours is that 0:02:20.120 --> 0:02:23.040 cuts are coming at the Whole Foods division, particularly. 0:02:22.520 --> 0:02:23.320 In corporate roles. 0:02:23.360 --> 0:02:26.399 We know about where Amazon's been trimming and that those 0:02:26.480 --> 0:02:29.160 layoffs and ounts have started, but it's interesting that Whole 0:02:29.160 --> 0:02:31.560 Foods as well is seeing some of that action. 0:02:31.840 --> 0:02:32.799 Let's go back to Tesla. 0:02:32.880 --> 0:02:35.560 Tesla's actually turned the corner in terms of its direction, 0:02:35.639 --> 0:02:37.880 its momentum in the last hour or so. We're hired 0:02:37.919 --> 0:02:40.880 by more than a percentage point, having been lower, markedly 0:02:40.919 --> 0:02:44.280 lower actually earlier in the session. The emphasis overnight is 0:02:44.280 --> 0:02:47.760 on price raises on model essenex here in the United States. 0:02:47.760 --> 0:02:49.720 But we go back to earnings a couple of days ago, 0:02:49.960 --> 0:02:52.880 the emphasis from elon musk more price cuts could come 0:02:53.080 --> 0:02:55.360 because this is a company CARO that is willing to 0:02:55.400 --> 0:02:58.600 sacrifice margin and profit to sustain that rate of growth 0:02:58.600 --> 0:03:01.000 that it's targeting fifty p sent or so on an 0:03:00.800 --> 0:03:02.359 annual basis going forward. 0:03:02.840 --> 0:03:04.560 Yeah, and I think we're going to be diving much 0:03:04.560 --> 0:03:07.520 more into the individual moves and into Tesla in a moment. 0:03:07.560 --> 0:03:09.960 But first let's talk about some of the geopolitics at 0:03:10.000 --> 0:03:13.320 play surrounding technology right now. President Biden, of course, he 0:03:13.320 --> 0:03:15.120 has been saying that he's going to crack down even 0:03:15.160 --> 0:03:17.840 more on US investment in key parts of China's economy, 0:03:17.840 --> 0:03:22.480 including in Chip's AI quantum computing protu. Moost Josh Wingrove 0:03:22.560 --> 0:03:25.840 covers this from DC and we sort of got hints 0:03:25.880 --> 0:03:28.080 a great scoop coming from some of your team in 0:03:28.160 --> 0:03:31.400 Washington about the move before the G seven summit. How 0:03:31.560 --> 0:03:34.120 likely are we like to see allies actually agree with this? 0:03:35.880 --> 0:03:38.320 Well, the move music isn't good right now that they're 0:03:38.360 --> 0:03:40.400 all going to join hands and proceed on this, but 0:03:40.440 --> 0:03:42.720 they might sort of give cover fire, if you will. 0:03:42.760 --> 0:03:45.080 And Biden kind of wants a coalition of the willing 0:03:45.080 --> 0:03:49.880 when it comes to sort of squeezing China's these sort 0:03:49.880 --> 0:03:52.360 of key sectors and squeezing investment in China. So even 0:03:52.360 --> 0:03:56.600 if for instance, Europe, Canada, Japan aren't enacting the same 0:03:56.640 --> 0:03:59.760 type of restrictions, if they give either an explicit or 0:04:00.040 --> 0:04:02.160 acid blessing, it looks like we'll see the US go 0:04:02.240 --> 0:04:05.480 ahead with this with his executive order. Now, this is 0:04:05.480 --> 0:04:08.440 our colleague Jenny Leonard reporting it. It will be sort 0:04:08.480 --> 0:04:11.280 of aimed at a few sectors, and it is preliminary. 0:04:11.360 --> 0:04:13.840 A the order itself is still in the works be 0:04:14.480 --> 0:04:17.520 it would then lead to regulations which take a long time, 0:04:17.720 --> 0:04:19.920 and you know, the devil is in the details. So 0:04:20.160 --> 0:04:22.760 this is still very early stages. But what it looks 0:04:22.800 --> 0:04:26.080 like is the US will proceed probably on its own 0:04:26.160 --> 0:04:29.400 right now, according to our colleagues reporting, with the sort 0:04:29.400 --> 0:04:31.480 of blessing of its allies. And this of course will 0:04:31.520 --> 0:04:35.440 further risk inflaming tensions between Beijing and Washington at a 0:04:35.480 --> 0:04:38.480 time where the sort of you know, wrangling over these 0:04:38.520 --> 0:04:40.760 economic restrictions has really been on the upswing. 0:04:41.640 --> 0:04:41.919 Josh. 0:04:42.120 --> 0:04:45.040 The logic from the administration, whether it is semiconductor or 0:04:45.160 --> 0:04:49.159 artificial intelligence related, has been the concern that China would 0:04:49.160 --> 0:04:53.720 apply that technology for military or security purposes, going against 0:04:53.720 --> 0:04:56.400 the security interests in the United States in the chip 0:04:56.440 --> 0:04:58.600 sect to what's interesting is you kind of have Taiwan 0:04:59.160 --> 0:05:02.560 caught right in the here TSMC, and that's actually an 0:05:02.600 --> 0:05:04.880 example where the United States has got allies on side 0:05:04.920 --> 0:05:08.839 the Netherlands Japan in terms of curbing chip exports to 0:05:09.240 --> 0:05:12.840 China through the Taiwan straight What is the latest there. 0:05:13.800 --> 0:05:16.720 Yeah, our colleagues are reporting that Taiwan is miffed a 0:05:16.800 --> 0:05:19.720 little bit that the US is sort of lumping Taiwan 0:05:19.880 --> 0:05:24.239 into its broader push, saying effectively, look, if you're trying 0:05:24.240 --> 0:05:27.680 to reorient away from Taiwan as part of a reorientation 0:05:27.839 --> 0:05:30.599 away from China, then in a way, you're kind of 0:05:30.600 --> 0:05:33.600 distancing yourself from Taiwan, which nominally the US is looking 0:05:33.640 --> 0:05:37.040 to support more broadly, and of course on the question 0:05:37.120 --> 0:05:39.160 of even when it would ever come to a question 0:05:39.200 --> 0:05:42.080 of military defense of Taiwan. So this is the kind 0:05:42.080 --> 0:05:44.520 of the really the big question. The US is trying 0:05:44.520 --> 0:05:47.120 to really steer those supply chains away, and that's got 0:05:47.120 --> 0:05:49.920 people's backs up in Taiwan who are saying, look, you know, 0:05:50.040 --> 0:05:53.360 like we were trying to be your friend here, maybe 0:05:53.400 --> 0:05:56.440 just don't don't lump us all in, don't paint with 0:05:56.440 --> 0:05:59.000 the same a big brush when it comes to dealing 0:05:59.080 --> 0:06:02.039 with these issues. His attention to the US is trying 0:06:02.040 --> 0:06:05.400 to do. Interestingly, in that story, they're particularly peeved at 0:06:05.800 --> 0:06:08.320 the Commerce Secretary Gina Raymando, who they think has been 0:06:08.360 --> 0:06:12.000 really ringing the bell on this one. And Raymondo has said, Look, 0:06:12.080 --> 0:06:14.039 the US is not looking to make one hundred percent 0:06:14.040 --> 0:06:16.479 of chips, for instance in the US, but they need 0:06:16.480 --> 0:06:19.960 to make more. It's not just that question of spying, sponage, 0:06:20.000 --> 0:06:23.560 military capability. It's also simply a question of availability. They 0:06:23.600 --> 0:06:25.080 don't want to be left without these things. 0:06:25.600 --> 0:06:27.920 Yes, certain the conversations I'm having around the world are 0:06:27.960 --> 0:06:30.359 about the concentration of supply chain in Asia and not 0:06:30.400 --> 0:06:33.479 necessarily bringing all on shoring one hundred percent to this nation. 0:06:33.600 --> 0:06:36.160 Joshuaen Gerry about a DC, Thank you so much. Look, 0:06:36.200 --> 0:06:40.120 another rival to open AI's chat GPT has hit the market. 0:06:40.200 --> 0:06:44.120 Stability AI just launched Stable LM, a chat bottel designed 0:06:44.120 --> 0:06:47.640 principally for researchers. But the model is already experiencing a 0:06:47.640 --> 0:06:50.599 couple of issues and answering simple prompts. 0:06:50.200 --> 0:06:53.000 Like what is five plus five? 0:06:53.440 --> 0:06:57.559 By meandering into a meditation about square inches and feet? 0:06:57.760 --> 0:06:58.680 What does all of that mean? 0:06:58.720 --> 0:07:03.280 Will STABILITYI not that it's growing pains with this early technology. 0:07:03.560 --> 0:07:07.240 We've also reported here at Bloomberg the company raising funds 0:07:07.240 --> 0:07:09.760 at a four billion dollar valuation according to sources. 0:07:09.840 --> 0:07:13.200 Let's get all of that with CEO Emmad Mustak for more. 0:07:13.640 --> 0:07:16.240 Immad, Welcome to the program, Caroline, and I have been 0:07:16.240 --> 0:07:20.360 looking forward to speaking with you. This model, it's targeted 0:07:20.360 --> 0:07:23.800 at researchers, right, it's a little bit nascent. What is 0:07:23.840 --> 0:07:25.680 it you think researchers will do with it? 0:07:27.280 --> 0:07:29.840 So I think that you have two paradigms here. You 0:07:29.880 --> 0:07:34.160 have the proprietary models, which includes open Ai, chatterpt and 0:07:34.200 --> 0:07:35.640 the rapt bit, Google, et cetera. And then you have 0:07:35.680 --> 0:07:38.520 the open paradigm where people are taking these models, they're 0:07:38.520 --> 0:07:42.360 experimenting with them, and they're doing wonderful things. So last 0:07:42.400 --> 0:07:47.000 August we collaborate and launch Stable Diffusion, this text image 0:07:47.000 --> 0:07:49.360 model that drove Lensa and one of these avatar apps. 0:07:49.400 --> 0:07:50.720 It was four of the top ten apps in the 0:07:50.720 --> 0:07:54.520 app Store, and the developer community there overtook Bitcoin in 0:07:54.520 --> 0:07:57.600 the theorem in cumulatative likes within two months. The ten 0:07:57.680 --> 0:08:00.000 years of that and it's become almost as popular as 0:07:59.880 --> 0:08:01.560 looks now in terms of. 0:08:01.480 --> 0:08:02.640 All these people working on it. 0:08:03.040 --> 0:08:05.960 So rather than basically take a gigantic model, put it 0:08:06.000 --> 0:08:08.360 out there a keep it secret, this approach is stead 0:08:08.400 --> 0:08:10.320 is iterative. You start with the seed and then you 0:08:10.360 --> 0:08:13.680 build it up, and the iteration feedback loop means it's 0:08:13.680 --> 0:08:15.800 going to get better and better live in front of 0:08:15.840 --> 0:08:19.360 people as the whole world yes and hacks on it, 0:08:19.400 --> 0:08:21.960 extends it, and it'll be super interesting to see what happens. 0:08:23.080 --> 0:08:25.239 M do you take us an interesting place quite early, 0:08:25.320 --> 0:08:27.800 which is everyone focused on the kind of billions of 0:08:27.800 --> 0:08:31.800 parameters that open ai is using for its llms. Actually 0:08:31.880 --> 0:08:33.600 a lot of what people are doing out there the 0:08:33.720 --> 0:08:37.719 inputs are based on a few one hundred million parameters 0:08:37.720 --> 0:08:40.199 at a lower scale, requiring less compute. 0:08:40.440 --> 0:08:41.319 There's still a question. 0:08:41.160 --> 0:08:43.720 About guard rails, and I'm interested in what guardrails you 0:08:43.720 --> 0:08:46.640 put in place for stable LM. 0:08:47.200 --> 0:08:49.360 So it was never our interest to build big models. 0:08:49.400 --> 0:08:51.079 This came from a lot of the labs wanted to 0:08:51.120 --> 0:08:54.480 build AGI generalized stages could do everything, so GPT four 0:08:54.520 --> 0:08:57.560 can pass the bar exam and GRE and GMT and 0:08:57.600 --> 0:08:58.360 everything like that. 0:08:58.720 --> 0:09:00.000 It's an amazing piece of software. 0:09:00.200 --> 0:09:02.600 But these big models are like really talented grads that 0:09:02.640 --> 0:09:05.319 occasionally go off their mets and they've fed jug food 0:09:05.320 --> 0:09:08.160 the whole internet because nobody's done the work on what 0:09:08.240 --> 0:09:11.079 is the actual data you need optimally for these models. 0:09:11.520 --> 0:09:13.320 What data do you need to feed these models so 0:09:13.360 --> 0:09:15.520 you can use them inside Bloomberg or a hedge fund 0:09:15.600 --> 0:09:18.560 or irregulated industry, so you have data transparency as well 0:09:18.559 --> 0:09:20.920 with these things. So the guardrails that we've put in 0:09:20.960 --> 0:09:23.840 place are trying to work with various parties. We've anounced 0:09:23.880 --> 0:09:26.240 Amazon as a partner in many mores coming on setting 0:09:26.280 --> 0:09:29.440 standards around open models because they are essential for private data. 0:09:29.960 --> 0:09:32.959 We have done things like in our previous image model 0:09:32.960 --> 0:09:35.280 that we released, we offered opt out and we're the 0:09:35.360 --> 0:09:37.199 only company in the world to do that, because what 0:09:37.240 --> 0:09:38.720 if artists don't want to be in the data sets 0:09:38.720 --> 0:09:41.680 and we offered full transparency around that allowed them to 0:09:41.679 --> 0:09:43.920 opt out. And I think these things will be interestingly 0:09:43.960 --> 0:09:48.480 important as you need standardization and industry self regulation at 0:09:48.480 --> 0:09:51.320 the big company level, at the anthropic open AI, cutting 0:09:51.440 --> 0:09:54.160 edge big model level, and at the open level. And 0:09:54.280 --> 0:09:57.920 we are looking specially there at the open level. That's 0:09:57.960 --> 0:10:00.280 why I kind of signed the FLI letter a few 0:10:00.280 --> 0:10:02.560 weeks ago, because I think now is the time for 0:10:02.600 --> 0:10:05.080 self regulation to take a pause and do this right. 0:10:05.400 --> 0:10:07.839 Yeah, because it's going around the world incredibly. 0:10:07.480 --> 0:10:12.720 Quickly, many will know stability AI well because of well 0:10:12.800 --> 0:10:16.000 already the image text to image that you've so put 0:10:16.000 --> 0:10:19.040 out their stable diffusion and you say you gave the 0:10:19.080 --> 0:10:22.040 opt out, but many it didn't stop some of the 0:10:22.120 --> 0:10:25.240 lawsuits coming your direction. How do you win hearts and 0:10:25.320 --> 0:10:28.760 minds of people that are worried about where the data 0:10:28.840 --> 0:10:31.959 is coming from and ultimately actually who owns it and 0:10:32.000 --> 0:10:33.120 who benefits from it. 0:10:34.520 --> 0:10:36.160 I think the way that you do this is you 0:10:36.200 --> 0:10:37.600 put it out in the open and have an open 0:10:37.600 --> 0:10:41.040 discussion around it. They're working with multiple governments now and 0:10:41.240 --> 0:10:42.640 helping them with national. 0:10:42.280 --> 0:10:43.920 Models as well with open data sets. 0:10:43.920 --> 0:10:45.880 And I think we should move away from web crawls 0:10:46.200 --> 0:10:48.600 and these are random things to having really high quality 0:10:48.679 --> 0:10:51.439 data that reflects the diversity of the world and is 0:10:51.520 --> 0:10:52.480 open and interpretive. 0:10:52.559 --> 0:10:53.920 That at Watertable and I. 0:10:53.880 --> 0:10:56.440 Said, there's no regulated industry or government that can use 0:10:56.480 --> 0:10:58.839 a close black box model. You need to have open 0:10:58.880 --> 0:11:01.000 models that come to the data that you own. So 0:11:01.040 --> 0:11:03.720 a lot of these companies, these really talented grads that 0:11:03.760 --> 0:11:07.520 are these models. Effectively, you know, they're like hiring for McKinsey, 0:11:07.600 --> 0:11:09.679 whereas you want to hire your own and have that 0:11:09.760 --> 0:11:11.880 because you can't give away your private knowledge. 0:11:12.040 --> 0:11:13.600 But again, there's so many standards and. 0:11:13.600 --> 0:11:15.480 Questions that need to be done around this, and we 0:11:15.559 --> 0:11:18.520 have to have those discussions now because the technology is 0:11:18.559 --> 0:11:19.680 literally going everywhere. 0:11:19.920 --> 0:11:21.400 I think it will be the number. 0:11:21.120 --> 0:11:23.760 One topic on Earning Scores and then definitely every government 0:11:23.800 --> 0:11:25.120 policy within the next few months. 0:11:25.200 --> 0:11:28.400 Do you have confidence that either you can self regulate 0:11:29.000 --> 0:11:33.079 or that some sort of multinational agreement can come to bear. 0:11:34.720 --> 0:11:35.600 I think it will be both. 0:11:35.679 --> 0:11:38.079 I think that you've got regulation, you have legislation, and 0:11:38.120 --> 0:11:41.160 you're seeing legislation come to bear, like Italy banning CHATCHEPT 0:11:41.320 --> 0:11:44.839 for example, and the EU looking at GDPR laws, whereas 0:11:44.880 --> 0:11:46.280 Japan has a very different regime. 0:11:46.280 --> 0:11:47.680 In the UK has to writing regime. 0:11:47.960 --> 0:11:49.920 One of the reasons I created Stability has a headflow 0:11:50.000 --> 0:11:52.800 manager before say to be on Alexis Show, was to 0:11:52.840 --> 0:11:56.000 try and make an entity that could help organize this 0:11:56.160 --> 0:11:58.959 that scale across the open area, even as this other 0:11:59.000 --> 0:12:02.040 regulation in the cut edge area with giant models and 0:12:02.120 --> 0:12:03.600 at the big company level as well. 0:12:05.360 --> 0:12:08.400 Am I Bloomberg's reported you're raising funds at a four 0:12:08.400 --> 0:12:09.440 billion dollar valuation. 0:12:09.559 --> 0:12:10.640 Did you close that round? 0:12:11.920 --> 0:12:13.520 I can't talk about things like that. 0:12:14.080 --> 0:12:16.120 We're the only company in the world that can build 0:12:16.160 --> 0:12:17.640 any model of any type for anyone. 0:12:18.320 --> 0:12:19.680 So it's a very interesting time. 0:12:21.040 --> 0:12:25.280 How about revenue generations, go ahead, current ad talk to 0:12:25.400 --> 0:12:26.680 us about how you make money. 0:12:28.400 --> 0:12:30.120 It's an open core model. So this is where I 0:12:30.160 --> 0:12:33.320 saw the arbitrage again the head trad manager past. What 0:12:33.440 --> 0:12:35.000 we're going to do is we're going to build the benchmark. 0:12:35.080 --> 0:12:36.840 We are the one of the largest putters of open 0:12:36.840 --> 0:12:39.679 source in the world, academia, others, because there's a big 0:12:39.679 --> 0:12:42.360 gap there with our supercompute. And then we take that 0:12:42.440 --> 0:12:44.240 and build on stable series of models that are the 0:12:44.280 --> 0:12:47.520 benchmark models that we build from scratch with the data sets. 0:12:47.559 --> 0:12:50.559 We have commercial partners that license it to financial variance, 0:12:50.559 --> 0:12:53.880 insurance variance, healthcare variance, and the language models, et cetera. 0:12:54.240 --> 0:12:57.240 And then national variants of that, and then through partners 0:12:57.280 --> 0:12:59.800 such as Amazon with Bedrock, you can go on Amazon, 0:13:00.120 --> 0:13:02.400 take that to your data and fully own your own model. 0:13:02.840 --> 0:13:03.800 And you know, we share on. 0:13:03.720 --> 0:13:07.160 The upside with our partners at Hyperscalar system integrators. And 0:13:07.200 --> 0:13:09.439 then for a series of companies that will announce soon, 0:13:09.840 --> 0:13:12.280 we're building them dedicated a item so they can keep 0:13:12.320 --> 0:13:15.280 on top of this. And that's an incredibly great business 0:13:15.280 --> 0:13:17.959 model because we offer stability at times has meant kind 0:13:17.960 --> 0:13:19.640 of chaos because no one knows what's going to happen. 0:13:20.920 --> 0:13:21.240 Am I do? 0:13:21.280 --> 0:13:25.160 You tweeted about AWS and Bedrock there was interesting How 0:13:25.240 --> 0:13:28.720 much of a differentiator is it to partner with tech 0:13:28.760 --> 0:13:32.400 companies that have mL specialized silicon and are working to 0:13:32.800 --> 0:13:35.120 improve the compute part of this process. 0:13:36.559 --> 0:13:39.079 I think it's a huge differentiator because you're moving from 0:13:39.200 --> 0:13:42.120 the GPU age to the asset age, just like you 0:13:42.120 --> 0:13:44.680 saw with bitcoin. So we work closely with Amazon on 0:13:44.720 --> 0:13:47.400 Inferentia too, which is an amazing diece of silicon that 0:13:47.440 --> 0:13:50.679 can run these workloads much much cheaper, and you'll see 0:13:50.679 --> 0:13:52.000 that being released an announcement. 0:13:52.000 --> 0:13:53.080 I think it's just John ga. 0:13:53.520 --> 0:13:56.800 The Bedrock service itself means that a company switches ours 0:13:56.800 --> 0:13:58.560 as eighteen months old, one hundred and eighty people. 0:13:58.679 --> 0:13:59.360 We do so much. 0:13:59.400 --> 0:14:00.920 We do all these models and all the types of 0:14:01.000 --> 0:14:04.320 video models, code models, language models. We can instead leverage 0:14:04.360 --> 0:14:08.480 that infrastructure. Take there's two one hundred thousand companies at 0:14:08.800 --> 0:14:12.280 stage Maker, and you know they can face Amazon. 0:14:11.920 --> 0:14:14.080 Instead of us because that's who they trust. 0:14:14.160 --> 0:14:15.760 And again, you want trust at a time like this, 0:14:15.920 --> 0:14:17.960 You want to have customs silicon that can run this 0:14:18.000 --> 0:14:20.840 incredibly cheaply, and you want standards around it. This is 0:14:20.880 --> 0:14:23.760 why our stable diffusion model is the first model on 0:14:23.880 --> 0:14:26.760 the Aerial Neural Engine at the start of December. Yeah, 0:14:26.840 --> 0:14:29.720 and so the customization optimization is going to be key. 0:14:30.000 --> 0:14:32.280 The key thing is distribution, because so many people want 0:14:32.320 --> 0:14:35.479 this and no company can scale fast enough and trustworthy 0:14:35.480 --> 0:14:38.000 manner enough to be able to support those slayers, et cetera. 0:14:38.480 --> 0:14:41.600 Stability AI CEO M. D. Mustak. Great to have some 0:14:41.640 --> 0:14:43.680 time with you. We thank you for working us into 0:14:43.680 --> 0:14:51.080 your calendar. It's busy, We appreciate it. 0:14:54.040 --> 0:14:56.720 Ev maker Tesla is raising the prices of its Model 0:14:56.840 --> 0:14:59.560 S and X in the United States. Comes after steep 0:14:59.560 --> 0:15:03.040 price earlier this year, which took a toll on profits, 0:15:03.080 --> 0:15:05.480 and of course it's had an impact weighing down the shares. 0:15:05.800 --> 0:15:08.840 It's just two days after test the lowered prices of 0:15:08.880 --> 0:15:11.160 some of its other models. We're tracking all the moves. 0:15:11.160 --> 0:15:14.520 Bloomberg Show and O came with us from Austin. The 0:15:14.640 --> 0:15:17.240 narrative does not match the action, right Sean, So tell 0:15:17.320 --> 0:15:19.000 us what we know about Model S and X in 0:15:19.040 --> 0:15:21.960 the US versus what Musk has been saying this week 0:15:22.000 --> 0:15:23.160 about strategy. 0:15:24.880 --> 0:15:26.920 Yeah, I mean, I think what we're seeing here is, 0:15:27.280 --> 0:15:31.120 you know, possibly a response to some of the pressure 0:15:31.200 --> 0:15:33.560 on the company as far as all of these price 0:15:33.600 --> 0:15:36.560 cuts that they've been sort of throwing at the Model 0:15:36.640 --> 0:15:38.440 Y and the Model three. But also I think it's 0:15:38.480 --> 0:15:41.720 important to remember that the first quarter of this year 0:15:41.880 --> 0:15:45.080 was one of the lower quarters of deliveries for the 0:15:45.080 --> 0:15:48.160 Model SNX over the last year and a half or so, 0:15:48.240 --> 0:15:53.360 basically since Tesla relaunched the sort of semi redesigned version 0:15:53.440 --> 0:15:56.480 of each of those cars about a year and a 0:15:56.480 --> 0:15:59.600 half ago. So this is you know that sort of 0:15:59.600 --> 0:16:02.920 almost counterintuitively, you think they would want to cut prices 0:16:02.960 --> 0:16:04.440 to increase the demand on those cars. 0:16:04.440 --> 0:16:06.480 But I think this is them saying, Hey, if. 0:16:06.320 --> 0:16:08.040 Our numbers are already kind of low on these cars, 0:16:08.120 --> 0:16:09.920 we might as well make back some of the margin here, 0:16:09.960 --> 0:16:12.320 even though it's obviously not going to make up, you know, 0:16:12.440 --> 0:16:14.760 for all of the money that they're losing by cutting 0:16:14.800 --> 0:16:16.280 some drastically on the Model Why. 0:16:16.680 --> 0:16:20.120 Yeah, this is where they've got to kind of square 0:16:20.160 --> 0:16:23.400 the circle. They have a very luxury end that actually 0:16:23.400 --> 0:16:25.480 is still way cheaper than it wasn't the start of 0:16:25.520 --> 0:16:27.320 the year, but it insured up by two or three 0:16:27.360 --> 0:16:29.920 percent on the day. But they're trying to keep the 0:16:30.000 --> 0:16:34.080 allure of that luxury number the xcs well at the 0:16:34.120 --> 0:16:37.040 same time making the threes and the ys just really 0:16:37.280 --> 0:16:40.040 mass scale here. Can they do that? Can they keep 0:16:40.080 --> 0:16:42.120 the allure of the higher end while still competing with 0:16:42.360 --> 0:16:43.800 Runner and foward ultimately? 0:16:45.720 --> 0:16:47.200 I mean, I think that's what we're really going to 0:16:47.240 --> 0:16:49.840 see this year. And you know, I think one of 0:16:49.840 --> 0:16:52.600 the things that's instructive when thinking about Tesla is that 0:16:53.120 --> 0:16:55.440 some of these other companies that are looking to try 0:16:55.440 --> 0:16:58.080 to replicate their success in the luxury space as far 0:16:58.080 --> 0:17:01.680 as the startups go, are actually really struggling. Lucid Motors 0:17:01.680 --> 0:17:03.840 has spent the last couple of months talking about how 0:17:04.520 --> 0:17:07.440 they're actually really having trouble creating demand and they're throwing 0:17:07.480 --> 0:17:10.840 more money into marketing to try to find customers for 0:17:10.920 --> 0:17:13.840 their expensive cars. And I think Tesla's always been kind 0:17:13.880 --> 0:17:17.080 of content ever since the model why especially took off 0:17:17.119 --> 0:17:19.800 in popularity, to let the S and X sort of 0:17:19.840 --> 0:17:22.920 be these sort of high end things that don't really 0:17:22.920 --> 0:17:27.040 push too foreign volume. They've started selling versions of those 0:17:27.920 --> 0:17:30.360 in China, these newer versions, and so I think they're 0:17:30.359 --> 0:17:32.360 going to find some homes for it, but I don't 0:17:32.359 --> 0:17:35.240 think they really think of it as as a big 0:17:35.359 --> 0:17:38.119 driver for them, And it just sort of remains this 0:17:38.200 --> 0:17:40.240 kind of like legacy halo thing that if you have 0:17:40.320 --> 0:17:43.040 the money, it's probably nice to have. But people seeing 0:17:43.080 --> 0:17:44.800 content with the Model Y and the Model three and 0:17:45.080 --> 0:17:49.320 Tesla seems content to push volume on those models and 0:17:49.560 --> 0:17:50.800 really just let that lead. 0:17:51.119 --> 0:17:54.560 Yeah, sure to Kaine trying to weave the narrative for us, 0:17:54.640 --> 0:17:56.640 as I'm sure you know Musk is trying to weave 0:17:56.640 --> 0:17:59.160 the sales as well. We thank you in other Musk 0:17:59.280 --> 0:18:01.840 related news to do it. Twitter stirring up all sorts 0:18:01.880 --> 0:18:04.919 of confusion after the social media platform remove those legacy 0:18:05.000 --> 0:18:08.040 check marks for verified accounts that aren't paying for its 0:18:08.040 --> 0:18:11.639 Twitter Blue program. Yet not everyone had to say goodbye 0:18:11.640 --> 0:18:13.960 to the blue checks? What was going on? Asia Counts 0:18:14.040 --> 0:18:15.600 joins us now right here in New York to talk 0:18:15.640 --> 0:18:20.159 us through it, and Asia seemingly Elon is paying for 0:18:20.240 --> 0:18:23.240 some blue check marks himself. How does that make sense? 0:18:23.400 --> 0:18:26.120 Yeah, it's a really weird situation. So, as you mentioned, 0:18:26.240 --> 0:18:28.879 legacy blue check marks went away in April twentieth. That 0:18:29.040 --> 0:18:32.960 means celebrities, athletes, anyone that's sort of notable under the 0:18:32.960 --> 0:18:36.640 old verification process lost their check mark. And so Musk 0:18:36.720 --> 0:18:38.359 has been very vocal about saying people are gonna have 0:18:38.359 --> 0:18:40.680 to pay for them. But then he decided to pay 0:18:40.720 --> 0:18:44.320 for a few celebrities, so Lebron, James, William Shatner and 0:18:44.320 --> 0:18:47.200 then Stephen King because they had been sort of vocal 0:18:47.280 --> 0:18:49.000 about not wanting to pay for the check So I 0:18:49.040 --> 0:18:50.920